Dhivehi and English/Final project - Revision history

Xluo1: /* RBMT */

2019-05-15T02:03:44Z

‎RBMT

Nmaini1: /* Improving the RBMT system */

2019-05-15T02:02:35Z

‎Improving the RBMT system

Nmaini1: /* Future Work */

2019-05-15T02:02:17Z

‎Future Work

Nmaini1: /* Dhivehi RBMT and ONMT */

2019-05-15T01:56:10Z

‎Dhivehi RBMT and ONMT

Nmaini1: /* Cross Validation for NMT */

2019-05-15T01:55:14Z

‎Cross Validation for NMT

Nmaini1: /* References */

2019-05-15T01:53:29Z

‎References

Nmaini1: /* Evaluation */

2019-05-15T01:52:54Z

‎Evaluation

Nmaini1: /* References */

2019-05-15T01:51:41Z

‎References

Nmaini1: /* References */

2019-05-15T01:48:12Z

‎References

Nmaini1: /* Dhivehi RBMT and ONMT */

2019-05-15T01:38:17Z

‎Dhivehi RBMT and ONMT

@@ Line 13: / Line 13: @@
 ===RBMT===
 Using Apertium, we studied the grammatical structure of Dhivehi and wrote various rules (phonological, syntactic, etc.) that map Dhivehi to English.
 ===NMT===

@@ Line 83: / Line 83: @@
 ====Improving the RBMT system====
-:* Implement more grammatical structures
+:* Implement more grammatical structures including:
 ::# Honorific System
 ::# Verb Tenses and Forms

← Older revision		Revision as of 02:02, 15 May 2019
Line 81:		Line 81:

	==Future Work==		==Future Work==
		+
		+	====Improving the RBMT system====
		+	:* Implement more grammatical structures
		+	::# Honorific System
		+	::# Verb Tenses and Forms
		+	::# Pronouns
		+
		+	====Improving the NMT system====
		+	:* Tweak hyper-parameters associated with the neural net such as learning rate, number of hidden layers, and others
		+	:* Expand size of parallel corpus

	==Github==		==Github==

← Older revision		Revision as of 01:56, 15 May 2019
Line 79:		Line 79:

	In short, if we had a large parallel corpus of modern text, then the NMT implementation would likely provide decent translation given enough training time.		In short, if we had a large parallel corpus of modern text, then the NMT implementation would likely provide decent translation given enough training time.
		+
		+	==Future Work==

	==Github==		==Github==

@@ Line 48: / Line 48: @@
 |}
-We used three-fold cross validation and found that the second model performed the best against its corresponding validation data set. We deemed it the best because the WER's of all the models were somewhat similar, but model two's PER was significantly lower. We can note that model 2's prediction feature more words overall than the other two models.
+We used three-fold cross validation and found that the second model performed the best against its corresponding validation data set. We deemed it the best because the WER's of all the models were somewhat similar, but model two's PER was significantly lower. We can note that model two's prediction feature more words overall than the other two models.
 ====RBMT and NMT====

@@ Line 85: / Line 85: @@
 ==References==
-Apertium - http://wiki.apertium.org/wiki/Main_Page
+[http://wiki.apertium.org/wiki/Main_Page Apertium]
-OpenNMT - https://opennmt.net
+[https://opennmt.net OpenNMT]
-NMT - http://en.wikipedia.org/wiki/Neural_machine_translation
+[http://en.wikipedia.org/wiki/Neural_machine_translation Neural Machine Translation]
 [[Category:Sp19 FinalProjects]]

@@ Line 72: / Line 72: @@
 |}
-We found that an NMT model (the Transformer [http://opennmt.net/OpenNMT-tf/package/opennmt.models.transformer.html] model) that trained for approximately thirty-nine hours performed better in-terms of WER and PER than the RBMT we worked on this semester. We must note, however, the data pooled together for training, validating, and testing the NMT system consisted primarily of biblical text. So, it is likely that the majority of the test data was from the Bible.
+We found that an NMT model (the [http://opennmt.net/OpenNMT-tf/package/opennmt.models.transformer.html Transformer] model) that trained for approximately thirty-nine hours performed better in-terms of WER and PER than the RBMT we worked on this semester. We must note, however, the data pooled together for training, validating, and testing the NMT system consisted primarily of biblical text. So, it is likely that the majority of the test data was from the Bible.
 ====Meaning of Results====
-These results tell us that the NMT implementation is very promising. If we were able to find a relatively inexpensive translator (perhaps through [https://www.translation-services-usa.com/languages/dhivehi.php]) and be able to build a large parallel corpus consisting of more modernized phrases, then we could have a more effective "modern-day" NMT system. The RBMT is more adaptable to modernized texts as we can tweak the individual mappings of words and grammar rules.
+These results tell us that the NMT implementation is very promising. If we were able to find a relatively inexpensive translator (perhaps through [https://www.translation-services-usa.com/languages/dhivehi.php Translation Services]) and be able to build a large parallel corpus consisting of more modernized phrases, then we could have a more effective "modern-day" NMT system. The RBMT is more adaptable to modernized texts as we can tweak the individual mappings of words and grammar rules.
 In short, if we had a large parallel corpus of modern text, then the NMT implementation would likely provide decent translation given enough training time.